Commit | Line | Data |
a0ed51b3 |
1 | /* utf8.c |
2 | * |
bc89e66f |
3 | * Copyright (c) 1998-2001, Larry Wall |
a0ed51b3 |
4 | * |
5 | * You may distribute under the terms of either the GNU General Public |
6 | * License or the Artistic License, as specified in the README file. |
7 | * |
8 | */ |
9 | |
10 | /* |
11 | * 'What a fix!' said Sam. 'That's the one place in all the lands we've ever |
12 | * heard of that we don't want to see any closer; and that's the one place |
13 | * we're trying to get to! And that's just where we can't get, nohow.' |
14 | * |
15 | * 'Well do I understand your speech,' he answered in the same language; |
16 | * 'yet few strangers do so. Why then do you not speak in the Common Tongue, |
17 | * as is the custom in the West, if you wish to be answered?' |
18 | * |
19 | * ...the travellers perceived that the floor was paved with stones of many |
20 | * hues; branching runes and strange devices intertwined beneath their feet. |
21 | */ |
22 | |
23 | #include "EXTERN.h" |
864dbfa3 |
24 | #define PERL_IN_UTF8_C |
a0ed51b3 |
25 | #include "perl.h" |
26 | |
27 | /* Unicode support */ |
28 | |
eebe1485 |
29 | /* |
30 | =for apidoc A|U8*|uv_to_utf8|U8 *d|UV uv |
31 | |
32 | Adds the UTF8 representation of the Unicode codepoint C<uv> to the end |
33 | of the string C<d>; C<d> should be have at least C<UTF8_MAXLEN+1> free |
34 | bytes available. The return value is the pointer to the byte after the |
35 | end of the new character. In other words, |
36 | |
37 | d = uv_to_utf8(d, uv); |
38 | |
39 | is the recommended Unicode-aware way of saying |
40 | |
41 | *(d++) = uv; |
42 | |
43 | =cut |
44 | */ |
45 | |
dfe13c55 |
46 | U8 * |
eebe1485 |
47 | Perl_uv_to_utf8(pTHX_ U8 *d, UV uv) |
a0ed51b3 |
48 | { |
49 | if (uv < 0x80) { |
50 | *d++ = uv; |
51 | return d; |
52 | } |
53 | if (uv < 0x800) { |
54 | *d++ = (( uv >> 6) | 0xc0); |
55 | *d++ = (( uv & 0x3f) | 0x80); |
56 | return d; |
57 | } |
58 | if (uv < 0x10000) { |
59 | *d++ = (( uv >> 12) | 0xe0); |
60 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
61 | *d++ = (( uv & 0x3f) | 0x80); |
62 | return d; |
63 | } |
64 | if (uv < 0x200000) { |
65 | *d++ = (( uv >> 18) | 0xf0); |
66 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
67 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
68 | *d++ = (( uv & 0x3f) | 0x80); |
69 | return d; |
70 | } |
71 | if (uv < 0x4000000) { |
72 | *d++ = (( uv >> 24) | 0xf8); |
73 | *d++ = (((uv >> 18) & 0x3f) | 0x80); |
74 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
75 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
76 | *d++ = (( uv & 0x3f) | 0x80); |
77 | return d; |
78 | } |
79 | if (uv < 0x80000000) { |
80 | *d++ = (( uv >> 30) | 0xfc); |
81 | *d++ = (((uv >> 24) & 0x3f) | 0x80); |
82 | *d++ = (((uv >> 18) & 0x3f) | 0x80); |
83 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
84 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
85 | *d++ = (( uv & 0x3f) | 0x80); |
86 | return d; |
87 | } |
6b8eaf93 |
88 | #ifdef HAS_QUAD |
d7578b48 |
89 | if (uv < UTF8_QUAD_MAX) |
a0ed51b3 |
90 | #endif |
91 | { |
92 | *d++ = 0xfe; /* Can't match U+FEFF! */ |
93 | *d++ = (((uv >> 30) & 0x3f) | 0x80); |
94 | *d++ = (((uv >> 24) & 0x3f) | 0x80); |
95 | *d++ = (((uv >> 18) & 0x3f) | 0x80); |
96 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
97 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
98 | *d++ = (( uv & 0x3f) | 0x80); |
99 | return d; |
100 | } |
6b8eaf93 |
101 | #ifdef HAS_QUAD |
a0ed51b3 |
102 | { |
103 | *d++ = 0xff; /* Can't match U+FFFE! */ |
3c77ea2b |
104 | *d++ = 0x80; /* 6 Reserved bits */ |
105 | *d++ = (((uv >> 60) & 0x0f) | 0x80); /* 2 Reserved bits */ |
106 | *d++ = (((uv >> 54) & 0x3f) | 0x80); |
107 | *d++ = (((uv >> 48) & 0x3f) | 0x80); |
108 | *d++ = (((uv >> 42) & 0x3f) | 0x80); |
a0ed51b3 |
109 | *d++ = (((uv >> 36) & 0x3f) | 0x80); |
110 | *d++ = (((uv >> 30) & 0x3f) | 0x80); |
111 | *d++ = (((uv >> 24) & 0x3f) | 0x80); |
112 | *d++ = (((uv >> 18) & 0x3f) | 0x80); |
113 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
114 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
115 | *d++ = (( uv & 0x3f) | 0x80); |
116 | return d; |
117 | } |
118 | #endif |
119 | } |
120 | |
eebe1485 |
121 | /* |
122 | =for apidoc A|STRLEN|is_utf8_char|U8 *s |
123 | |
124 | Tests if some arbitrary number of bytes begins in a valid UTF-8 character. |
125 | The actual number of bytes in the UTF-8 character will be returned if it |
126 | is valid, otherwise 0. |
127 | |
128 | =cut |
129 | */ |
067a85ef |
130 | STRLEN |
386d01d6 |
131 | Perl_is_utf8_char(pTHX_ U8 *s) |
132 | { |
133 | U8 u = *s; |
067a85ef |
134 | STRLEN slen, len; |
135 | UV uv, ouv; |
386d01d6 |
136 | |
60006e79 |
137 | if (UTF8_IS_ASCII(u)) |
386d01d6 |
138 | return 1; |
139 | |
60006e79 |
140 | if (!UTF8_IS_START(u)) |
386d01d6 |
141 | return 0; |
142 | |
9f07fdcd |
143 | len = UTF8SKIP(s); |
386d01d6 |
144 | |
60006e79 |
145 | if (len < 2 || !UTF8_IS_CONTINUATION(s[1])) |
067a85ef |
146 | return 0; |
147 | |
386d01d6 |
148 | slen = len - 1; |
149 | s++; |
067a85ef |
150 | uv = u; |
151 | ouv = uv; |
386d01d6 |
152 | while (slen--) { |
60006e79 |
153 | if (!UTF8_IS_CONTINUATION(*s)) |
386d01d6 |
154 | return 0; |
8850bf83 |
155 | uv = UTF8_ACCUMULATE(uv, *s); |
067a85ef |
156 | if (uv < ouv) |
157 | return 0; |
158 | ouv = uv; |
386d01d6 |
159 | s++; |
160 | } |
067a85ef |
161 | |
5bbb0b5a |
162 | if (UNISKIP(uv) < len) |
067a85ef |
163 | return 0; |
164 | |
386d01d6 |
165 | return len; |
166 | } |
167 | |
6662521e |
168 | /* |
eebe1485 |
169 | =for apidoc A|bool|is_utf8_string|U8 *s|STRLEN len |
6662521e |
170 | |
171 | Returns true if first C<len> bytes of the given string form valid a UTF8 |
172 | string, false otherwise. |
173 | |
174 | =cut |
175 | */ |
176 | |
8e84507e |
177 | bool |
6662521e |
178 | Perl_is_utf8_string(pTHX_ U8 *s, STRLEN len) |
179 | { |
067a85ef |
180 | U8* x = s; |
1aa99e6b |
181 | U8* send; |
067a85ef |
182 | STRLEN c; |
183 | |
1aa99e6b |
184 | if (!len) |
6cd5fe39 |
185 | len = strlen((char *)s); |
1aa99e6b |
186 | send = s + len; |
187 | |
6662521e |
188 | while (x < send) { |
189 | c = is_utf8_char(x); |
067a85ef |
190 | if (!c) |
191 | return FALSE; |
6662521e |
192 | x += c; |
6662521e |
193 | } |
60006e79 |
194 | if (x != send) |
195 | return FALSE; |
067a85ef |
196 | |
197 | return TRUE; |
6662521e |
198 | } |
199 | |
67e989fb |
200 | /* |
eebe1485 |
201 | =for apidoc A|U8* s|utf8_to_uv|STRLEN curlen|STRLEN *retlen|U32 flags |
67e989fb |
202 | |
203 | Returns the character value of the first character in the string C<s> |
ba210ebe |
204 | which is assumed to be in UTF8 encoding and no longer than C<curlen>; |
7df053ec |
205 | C<retlen> will be set to the length, in bytes, of that character. |
67e989fb |
206 | |
207 | If C<s> does not point to a well-formed UTF8 character, the behaviour |
dcad2880 |
208 | is dependent on the value of C<flags>: if it contains UTF8_CHECK_ONLY, |
209 | it is assumed that the caller will raise a warning, and this function |
28d3d195 |
210 | will silently just set C<retlen> to C<-1> and return zero. If the |
211 | C<flags> does not contain UTF8_CHECK_ONLY, warnings about |
212 | malformations will be given, C<retlen> will be set to the expected |
213 | length of the UTF-8 character in bytes, and zero will be returned. |
214 | |
215 | The C<flags> can also contain various flags to allow deviations from |
216 | the strict UTF-8 encoding (see F<utf8.h>). |
67e989fb |
217 | |
dcad2880 |
218 | =cut */ |
67e989fb |
219 | |
a0ed51b3 |
220 | UV |
dcad2880 |
221 | Perl_utf8_to_uv(pTHX_ U8* s, STRLEN curlen, STRLEN* retlen, U32 flags) |
a0ed51b3 |
222 | { |
ba210ebe |
223 | UV uv = *s, ouv; |
224 | STRLEN len = 1; |
7bf1b6bb |
225 | #ifdef EBCDIC |
226 | bool dowarn = 0; |
227 | #else |
ba210ebe |
228 | bool dowarn = ckWARN_d(WARN_UTF8); |
7bf1b6bb |
229 | #endif |
ba210ebe |
230 | STRLEN expectlen = 0; |
a0dbb045 |
231 | U32 warning = 0; |
232 | |
233 | /* This list is a superset of the UTF8_ALLOW_XXX. */ |
234 | |
235 | #define UTF8_WARN_EMPTY 1 |
236 | #define UTF8_WARN_CONTINUATION 2 |
237 | #define UTF8_WARN_NON_CONTINUATION 3 |
238 | #define UTF8_WARN_FE_FF 4 |
239 | #define UTF8_WARN_SHORT 5 |
240 | #define UTF8_WARN_OVERFLOW 6 |
241 | #define UTF8_WARN_SURROGATE 7 |
242 | #define UTF8_WARN_BOM 8 |
243 | #define UTF8_WARN_LONG 9 |
244 | #define UTF8_WARN_FFFF 10 |
245 | |
246 | if (curlen == 0 && |
247 | !(flags & UTF8_ALLOW_EMPTY)) { |
248 | warning = UTF8_WARN_EMPTY; |
0c443dc2 |
249 | goto malformed; |
250 | } |
251 | |
421a8bf2 |
252 | if (UTF8_IS_ASCII(uv)) { |
a0ed51b3 |
253 | if (retlen) |
254 | *retlen = 1; |
255 | return *s; |
256 | } |
67e989fb |
257 | |
421a8bf2 |
258 | if (UTF8_IS_CONTINUATION(uv) && |
fcc8fcf6 |
259 | !(flags & UTF8_ALLOW_CONTINUATION)) { |
a0dbb045 |
260 | warning = UTF8_WARN_CONTINUATION; |
ba210ebe |
261 | goto malformed; |
262 | } |
263 | |
421a8bf2 |
264 | if (UTF8_IS_START(uv) && curlen > 1 && !UTF8_IS_CONTINUATION(s[1]) && |
fcc8fcf6 |
265 | !(flags & UTF8_ALLOW_NON_CONTINUATION)) { |
a0dbb045 |
266 | warning = UTF8_WARN_NON_CONTINUATION; |
ba210ebe |
267 | goto malformed; |
268 | } |
fcc8fcf6 |
269 | |
270 | if ((uv == 0xfe || uv == 0xff) && |
271 | !(flags & UTF8_ALLOW_FE_FF)) { |
a0dbb045 |
272 | warning = UTF8_WARN_FE_FF; |
ba210ebe |
273 | goto malformed; |
a0ed51b3 |
274 | } |
fcc8fcf6 |
275 | |
ba210ebe |
276 | if (!(uv & 0x20)) { len = 2; uv &= 0x1f; } |
277 | else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; } |
278 | else if (!(uv & 0x08)) { len = 4; uv &= 0x07; } |
279 | else if (!(uv & 0x04)) { len = 5; uv &= 0x03; } |
280 | else if (!(uv & 0x02)) { len = 6; uv &= 0x01; } |
281 | else if (!(uv & 0x01)) { len = 7; uv = 0; } |
3c77ea2b |
282 | else { len = 13; uv = 0; } /* whoa! */ |
fcc8fcf6 |
283 | |
a0ed51b3 |
284 | if (retlen) |
285 | *retlen = len; |
ba210ebe |
286 | |
287 | expectlen = len; |
288 | |
fcc8fcf6 |
289 | if ((curlen < expectlen) && |
290 | !(flags & UTF8_ALLOW_SHORT)) { |
a0dbb045 |
291 | warning = UTF8_WARN_SHORT; |
ba210ebe |
292 | goto malformed; |
293 | } |
294 | |
295 | len--; |
a0ed51b3 |
296 | s++; |
ba210ebe |
297 | ouv = uv; |
298 | |
a0ed51b3 |
299 | while (len--) { |
421a8bf2 |
300 | if (!UTF8_IS_CONTINUATION(*s) && |
301 | !(flags & UTF8_ALLOW_NON_CONTINUATION)) { |
a0dbb045 |
302 | s--; |
303 | warning = UTF8_WARN_NON_CONTINUATION; |
ba210ebe |
304 | goto malformed; |
a0ed51b3 |
305 | } |
306 | else |
8850bf83 |
307 | uv = UTF8_ACCUMULATE(uv, *s); |
a0dbb045 |
308 | if (!(uv > ouv)) { |
309 | /* These cannot be allowed. */ |
310 | if (uv == ouv) { |
311 | if (!(flags & UTF8_ALLOW_LONG)) { |
312 | warning = UTF8_WARN_LONG; |
313 | goto malformed; |
314 | } |
315 | } |
316 | else { /* uv < ouv */ |
317 | /* This cannot be allowed. */ |
318 | warning = UTF8_WARN_OVERFLOW; |
319 | goto malformed; |
320 | } |
ba210ebe |
321 | } |
322 | s++; |
323 | ouv = uv; |
324 | } |
325 | |
421a8bf2 |
326 | if (UNICODE_IS_SURROGATE(uv) && |
fcc8fcf6 |
327 | !(flags & UTF8_ALLOW_SURROGATE)) { |
a0dbb045 |
328 | warning = UTF8_WARN_SURROGATE; |
ba210ebe |
329 | goto malformed; |
421a8bf2 |
330 | } else if (UNICODE_IS_BYTE_ORDER_MARK(uv) && |
fcc8fcf6 |
331 | !(flags & UTF8_ALLOW_BOM)) { |
a0dbb045 |
332 | warning = UTF8_WARN_BOM; |
ba210ebe |
333 | goto malformed; |
fcc8fcf6 |
334 | } else if ((expectlen > UNISKIP(uv)) && |
335 | !(flags & UTF8_ALLOW_LONG)) { |
a0dbb045 |
336 | warning = UTF8_WARN_LONG; |
ba210ebe |
337 | goto malformed; |
421a8bf2 |
338 | } else if (UNICODE_IS_ILLEGAL(uv) && |
a9917092 |
339 | !(flags & UTF8_ALLOW_FFFF)) { |
a0dbb045 |
340 | warning = UTF8_WARN_FFFF; |
a9917092 |
341 | goto malformed; |
a0ed51b3 |
342 | } |
ba210ebe |
343 | |
a0ed51b3 |
344 | return uv; |
ba210ebe |
345 | |
346 | malformed: |
347 | |
fcc8fcf6 |
348 | if (flags & UTF8_CHECK_ONLY) { |
ba210ebe |
349 | if (retlen) |
cc366d4b |
350 | *retlen = -1; |
ba210ebe |
351 | return 0; |
352 | } |
353 | |
a0dbb045 |
354 | if (dowarn) { |
355 | SV* sv = sv_2mortal(newSVpv("Malformed UTF-8 character ", 0)); |
356 | |
357 | switch (warning) { |
358 | case 0: /* Intentionally empty. */ break; |
359 | case UTF8_WARN_EMPTY: |
360 | Perl_sv_catpvf(aTHX_ sv, "(empty string)"); |
361 | break; |
362 | case UTF8_WARN_CONTINUATION: |
363 | Perl_sv_catpvf(aTHX_ sv, "(unexpected continuation byte 0x%02"UVxf")", uv); |
364 | break; |
365 | case UTF8_WARN_NON_CONTINUATION: |
366 | Perl_sv_catpvf(aTHX_ sv, "(unexpected non-continuation byte 0x%02"UVxf" after start byte 0x%02"UVxf")", |
367 | (UV)s[1], uv); |
368 | break; |
369 | case UTF8_WARN_FE_FF: |
370 | Perl_sv_catpvf(aTHX_ sv, "(byte 0x%02"UVxf")", uv); |
371 | break; |
372 | case UTF8_WARN_SHORT: |
373 | Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d)", |
374 | curlen, curlen == 1 ? "" : "s", expectlen); |
375 | break; |
376 | case UTF8_WARN_OVERFLOW: |
377 | Perl_sv_catpvf(aTHX_ sv, "(overflow at 0x%"UVxf", byte 0x%02x)", |
378 | ouv, *s); |
379 | break; |
380 | case UTF8_WARN_SURROGATE: |
381 | Perl_sv_catpvf(aTHX_ sv, "(UTF-16 surrogate 0x%04"UVxf")", uv); |
382 | break; |
383 | case UTF8_WARN_BOM: |
384 | Perl_sv_catpvf(aTHX_ sv, "(byte order mark 0x%04"UVxf")", uv); |
385 | break; |
386 | case UTF8_WARN_LONG: |
387 | Perl_sv_catpvf(aTHX_ sv, "(%d byte%s, need %d)", |
388 | expectlen, expectlen == 1 ? "": "s", UNISKIP(uv)); |
389 | break; |
390 | case UTF8_WARN_FFFF: |
391 | Perl_sv_catpvf(aTHX_ sv, "(character 0x%04"UVxf")", uv); |
392 | break; |
393 | default: |
394 | Perl_sv_catpvf(aTHX_ sv, "(unknown reason)"); |
395 | break; |
396 | } |
397 | |
398 | if (warning) { |
399 | char *s = SvPVX(sv); |
400 | |
401 | if (PL_op) |
402 | Perl_warner(aTHX_ WARN_UTF8, |
403 | "%s in %s", s, PL_op_desc[PL_op->op_type]); |
404 | else |
405 | Perl_warner(aTHX_ WARN_UTF8, "%s", s); |
406 | } |
407 | } |
408 | |
ba210ebe |
409 | if (retlen) |
28d3d195 |
410 | *retlen = expectlen ? expectlen : len; |
ba210ebe |
411 | |
28d3d195 |
412 | return 0; |
a0ed51b3 |
413 | } |
414 | |
8e84507e |
415 | /* |
eebe1485 |
416 | =for apidoc A|U8* s|utf8_to_uv_simple|STRLEN *retlen |
8e84507e |
417 | |
418 | Returns the character value of the first character in the string C<s> |
419 | which is assumed to be in UTF8 encoding; C<retlen> will be set to the |
7df053ec |
420 | length, in bytes, of that character. |
8e84507e |
421 | |
ba210ebe |
422 | If C<s> does not point to a well-formed UTF8 character, zero is |
423 | returned and retlen is set, if possible, to -1. |
8e84507e |
424 | |
425 | =cut |
426 | */ |
427 | |
428 | UV |
dcad2880 |
429 | Perl_utf8_to_uv_simple(pTHX_ U8* s, STRLEN* retlen) |
8e84507e |
430 | { |
2e4dc9fc |
431 | return Perl_utf8_to_uv(aTHX_ s, UTF8_MAXLEN, retlen, 0); |
8e84507e |
432 | } |
433 | |
b76347f2 |
434 | /* |
eebe1485 |
435 | =for apidoc A|STRLEN|utf8_length|U8* s|U8 *e |
b76347f2 |
436 | |
437 | Return the length of the UTF-8 char encoded string C<s> in characters. |
02eb7b47 |
438 | Stops at C<e> (inclusive). If C<e E<lt> s> or if the scan would end |
439 | up past C<e>, croaks. |
b76347f2 |
440 | |
441 | =cut |
442 | */ |
443 | |
444 | STRLEN |
445 | Perl_utf8_length(pTHX_ U8* s, U8* e) |
446 | { |
447 | STRLEN len = 0; |
448 | |
8850bf83 |
449 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g. |
450 | * the bitops (especially ~) can create illegal UTF-8. |
451 | * In other words: in Perl UTF-8 is not just for Unicode. */ |
452 | |
b76347f2 |
453 | if (e < s) |
02eb7b47 |
454 | Perl_croak(aTHX_ "panic: utf8_length: unexpected end"); |
b76347f2 |
455 | while (s < e) { |
02eb7b47 |
456 | U8 t = UTF8SKIP(s); |
b76347f2 |
457 | |
458 | if (e - s < t) |
02eb7b47 |
459 | Perl_croak(aTHX_ "panic: utf8_length: unaligned end"); |
b76347f2 |
460 | s += t; |
461 | len++; |
462 | } |
463 | |
464 | return len; |
465 | } |
466 | |
b06226ff |
467 | /* |
eebe1485 |
468 | =for apidoc A|IV|utf8_distance|U8 *a|U8 *b |
b06226ff |
469 | |
470 | Returns the number of UTF8 characters between the UTF-8 pointers C<a> |
471 | and C<b>. |
472 | |
473 | WARNING: use only if you *know* that the pointers point inside the |
474 | same UTF-8 buffer. |
475 | |
476 | =cut */ |
a0ed51b3 |
477 | |
02eb7b47 |
478 | IV |
864dbfa3 |
479 | Perl_utf8_distance(pTHX_ U8 *a, U8 *b) |
a0ed51b3 |
480 | { |
02eb7b47 |
481 | IV off = 0; |
482 | |
8850bf83 |
483 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g. |
484 | * the bitops (especially ~) can create illegal UTF-8. |
485 | * In other words: in Perl UTF-8 is not just for Unicode. */ |
486 | |
a0ed51b3 |
487 | if (a < b) { |
488 | while (a < b) { |
02eb7b47 |
489 | U8 c = UTF8SKIP(a); |
490 | |
491 | if (b - a < c) |
492 | Perl_croak(aTHX_ "panic: utf8_distance: unaligned end"); |
493 | a += c; |
a0ed51b3 |
494 | off--; |
495 | } |
496 | } |
497 | else { |
498 | while (b < a) { |
02eb7b47 |
499 | U8 c = UTF8SKIP(b); |
500 | |
501 | if (a - b < c) |
502 | Perl_croak(aTHX_ "panic: utf8_distance: unaligned end"); |
503 | b += c; |
a0ed51b3 |
504 | off++; |
505 | } |
506 | } |
02eb7b47 |
507 | |
a0ed51b3 |
508 | return off; |
509 | } |
510 | |
b06226ff |
511 | /* |
eebe1485 |
512 | =for apidoc A|U8*|utf8_hop|U8 *s|I32 off |
b06226ff |
513 | |
8850bf83 |
514 | Return the UTF-8 pointer C<s> displaced by C<off> characters, either |
515 | forward or backward. |
b06226ff |
516 | |
517 | WARNING: do not use the following unless you *know* C<off> is within |
8850bf83 |
518 | the UTF-8 data pointed to by C<s> *and* that on entry C<s> is aligned |
519 | on the first byte of character or just after the last byte of a character. |
b06226ff |
520 | |
521 | =cut */ |
a0ed51b3 |
522 | |
523 | U8 * |
864dbfa3 |
524 | Perl_utf8_hop(pTHX_ U8 *s, I32 off) |
a0ed51b3 |
525 | { |
8850bf83 |
526 | /* Note: cannot use UTF8_IS_...() too eagerly here since e.g |
527 | * the bitops (especially ~) can create illegal UTF-8. |
528 | * In other words: in Perl UTF-8 is not just for Unicode. */ |
529 | |
a0ed51b3 |
530 | if (off >= 0) { |
531 | while (off--) |
532 | s += UTF8SKIP(s); |
533 | } |
534 | else { |
535 | while (off++) { |
536 | s--; |
8850bf83 |
537 | while (UTF8_IS_CONTINUATION(*s)) |
538 | s--; |
a0ed51b3 |
539 | } |
540 | } |
541 | return s; |
542 | } |
543 | |
6940069f |
544 | /* |
eebe1485 |
545 | =for apidoc A|U8 *|utf8_to_bytes|U8 *s|STRLEN *len |
6940069f |
546 | |
246fae53 |
547 | Converts a string C<s> of length C<len> from UTF8 into byte encoding. |
548 | Unlike C<bytes_to_utf8>, this over-writes the original string, and |
549 | updates len to contain the new length. |
67e989fb |
550 | Returns zero on failure, setting C<len> to -1. |
6940069f |
551 | |
552 | =cut |
553 | */ |
554 | |
555 | U8 * |
246fae53 |
556 | Perl_utf8_to_bytes(pTHX_ U8* s, STRLEN *len) |
6940069f |
557 | { |
6940069f |
558 | U8 *send; |
559 | U8 *d; |
dcad2880 |
560 | U8 *save = s; |
246fae53 |
561 | |
562 | /* ensure valid UTF8 and chars < 256 before updating string */ |
dcad2880 |
563 | for (send = s + *len; s < send; ) { |
564 | U8 c = *s++; |
565 | |
9f9ab905 |
566 | if (c >= 0x80 && |
dcad2880 |
567 | ((s >= send) || |
568 | ((*s++ & 0xc0) != 0x80) || ((c & 0xfe) != 0xc2))) { |
569 | *len = -1; |
570 | return 0; |
571 | } |
246fae53 |
572 | } |
dcad2880 |
573 | |
574 | d = s = save; |
6940069f |
575 | while (s < send) { |
ed646e6e |
576 | STRLEN ulen; |
577 | *d++ = (U8)utf8_to_uv_simple(s, &ulen); |
578 | s += ulen; |
6940069f |
579 | } |
580 | *d = '\0'; |
246fae53 |
581 | *len = d - save; |
6940069f |
582 | return save; |
583 | } |
584 | |
585 | /* |
eebe1485 |
586 | =for apidoc A|U8 *|bytes_to_utf8|U8 *s|STRLEN *len |
6940069f |
587 | |
588 | Converts a string C<s> of length C<len> from ASCII into UTF8 encoding. |
6662521e |
589 | Returns a pointer to the newly-created string, and sets C<len> to |
590 | reflect the new length. |
6940069f |
591 | |
497711e7 |
592 | =cut |
6940069f |
593 | */ |
594 | |
595 | U8* |
6662521e |
596 | Perl_bytes_to_utf8(pTHX_ U8* s, STRLEN *len) |
6940069f |
597 | { |
6940069f |
598 | U8 *send; |
599 | U8 *d; |
600 | U8 *dst; |
6662521e |
601 | send = s + (*len); |
6940069f |
602 | |
6662521e |
603 | Newz(801, d, (*len) * 2 + 1, U8); |
6940069f |
604 | dst = d; |
605 | |
606 | while (s < send) { |
607 | if (*s < 0x80) |
608 | *d++ = *s++; |
609 | else { |
610 | UV uv = *s++; |
611 | *d++ = (( uv >> 6) | 0xc0); |
612 | *d++ = (( uv & 0x3f) | 0x80); |
613 | } |
614 | } |
615 | *d = '\0'; |
6662521e |
616 | *len = d-dst; |
6940069f |
617 | return dst; |
618 | } |
619 | |
a0ed51b3 |
620 | /* |
dea0fc0b |
621 | * Convert native (big-endian) or reversed (little-endian) UTF-16 to UTF-8. |
a0ed51b3 |
622 | * |
623 | * Destination must be pre-extended to 3/2 source. Do not use in-place. |
624 | * We optimize for native, for obvious reasons. */ |
625 | |
626 | U8* |
dea0fc0b |
627 | Perl_utf16_to_utf8(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 |
628 | { |
dea0fc0b |
629 | U8* pend; |
630 | U8* dstart = d; |
631 | |
632 | if (bytelen & 1) |
a7867d0a |
633 | Perl_croak(aTHX_ "panic: utf16_to_utf8: odd bytelen"); |
dea0fc0b |
634 | |
635 | pend = p + bytelen; |
636 | |
a0ed51b3 |
637 | while (p < pend) { |
dea0fc0b |
638 | UV uv = (p[0] << 8) + p[1]; /* UTF-16BE */ |
639 | p += 2; |
a0ed51b3 |
640 | if (uv < 0x80) { |
641 | *d++ = uv; |
642 | continue; |
643 | } |
644 | if (uv < 0x800) { |
645 | *d++ = (( uv >> 6) | 0xc0); |
646 | *d++ = (( uv & 0x3f) | 0x80); |
647 | continue; |
648 | } |
649 | if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */ |
dea0fc0b |
650 | UV low = *p++; |
651 | if (low < 0xdc00 || low >= 0xdfff) |
652 | Perl_croak(aTHX_ "Malformed UTF-16 surrogate"); |
a0ed51b3 |
653 | uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000; |
654 | } |
655 | if (uv < 0x10000) { |
656 | *d++ = (( uv >> 12) | 0xe0); |
657 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
658 | *d++ = (( uv & 0x3f) | 0x80); |
659 | continue; |
660 | } |
661 | else { |
662 | *d++ = (( uv >> 18) | 0xf0); |
663 | *d++ = (((uv >> 12) & 0x3f) | 0x80); |
664 | *d++ = (((uv >> 6) & 0x3f) | 0x80); |
665 | *d++ = (( uv & 0x3f) | 0x80); |
666 | continue; |
667 | } |
668 | } |
dea0fc0b |
669 | *newlen = d - dstart; |
a0ed51b3 |
670 | return d; |
671 | } |
672 | |
673 | /* Note: this one is slightly destructive of the source. */ |
674 | |
675 | U8* |
dea0fc0b |
676 | Perl_utf16_to_utf8_reversed(pTHX_ U8* p, U8* d, I32 bytelen, I32 *newlen) |
a0ed51b3 |
677 | { |
678 | U8* s = (U8*)p; |
679 | U8* send = s + bytelen; |
680 | while (s < send) { |
681 | U8 tmp = s[0]; |
682 | s[0] = s[1]; |
683 | s[1] = tmp; |
684 | s += 2; |
685 | } |
dea0fc0b |
686 | return utf16_to_utf8(p, d, bytelen, newlen); |
a0ed51b3 |
687 | } |
688 | |
689 | /* for now these are all defined (inefficiently) in terms of the utf8 versions */ |
690 | |
691 | bool |
864dbfa3 |
692 | Perl_is_uni_alnum(pTHX_ U32 c) |
a0ed51b3 |
693 | { |
ad391ad9 |
694 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
695 | uv_to_utf8(tmpbuf, (UV)c); |
696 | return is_utf8_alnum(tmpbuf); |
697 | } |
698 | |
699 | bool |
b8c5462f |
700 | Perl_is_uni_alnumc(pTHX_ U32 c) |
701 | { |
ad391ad9 |
702 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
703 | uv_to_utf8(tmpbuf, (UV)c); |
704 | return is_utf8_alnumc(tmpbuf); |
705 | } |
706 | |
707 | bool |
864dbfa3 |
708 | Perl_is_uni_idfirst(pTHX_ U32 c) |
a0ed51b3 |
709 | { |
ad391ad9 |
710 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
711 | uv_to_utf8(tmpbuf, (UV)c); |
712 | return is_utf8_idfirst(tmpbuf); |
713 | } |
714 | |
715 | bool |
864dbfa3 |
716 | Perl_is_uni_alpha(pTHX_ U32 c) |
a0ed51b3 |
717 | { |
ad391ad9 |
718 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
719 | uv_to_utf8(tmpbuf, (UV)c); |
720 | return is_utf8_alpha(tmpbuf); |
721 | } |
722 | |
723 | bool |
4d61ec05 |
724 | Perl_is_uni_ascii(pTHX_ U32 c) |
725 | { |
ad391ad9 |
726 | U8 tmpbuf[UTF8_MAXLEN+1]; |
4d61ec05 |
727 | uv_to_utf8(tmpbuf, (UV)c); |
728 | return is_utf8_ascii(tmpbuf); |
729 | } |
730 | |
731 | bool |
864dbfa3 |
732 | Perl_is_uni_space(pTHX_ U32 c) |
a0ed51b3 |
733 | { |
ad391ad9 |
734 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
735 | uv_to_utf8(tmpbuf, (UV)c); |
736 | return is_utf8_space(tmpbuf); |
737 | } |
738 | |
739 | bool |
864dbfa3 |
740 | Perl_is_uni_digit(pTHX_ U32 c) |
a0ed51b3 |
741 | { |
ad391ad9 |
742 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
743 | uv_to_utf8(tmpbuf, (UV)c); |
744 | return is_utf8_digit(tmpbuf); |
745 | } |
746 | |
747 | bool |
864dbfa3 |
748 | Perl_is_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
749 | { |
ad391ad9 |
750 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
751 | uv_to_utf8(tmpbuf, (UV)c); |
752 | return is_utf8_upper(tmpbuf); |
753 | } |
754 | |
755 | bool |
864dbfa3 |
756 | Perl_is_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
757 | { |
ad391ad9 |
758 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
759 | uv_to_utf8(tmpbuf, (UV)c); |
760 | return is_utf8_lower(tmpbuf); |
761 | } |
762 | |
763 | bool |
b8c5462f |
764 | Perl_is_uni_cntrl(pTHX_ U32 c) |
765 | { |
ad391ad9 |
766 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
767 | uv_to_utf8(tmpbuf, (UV)c); |
768 | return is_utf8_cntrl(tmpbuf); |
769 | } |
770 | |
771 | bool |
772 | Perl_is_uni_graph(pTHX_ U32 c) |
773 | { |
ad391ad9 |
774 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
775 | uv_to_utf8(tmpbuf, (UV)c); |
776 | return is_utf8_graph(tmpbuf); |
777 | } |
778 | |
779 | bool |
864dbfa3 |
780 | Perl_is_uni_print(pTHX_ U32 c) |
a0ed51b3 |
781 | { |
ad391ad9 |
782 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
783 | uv_to_utf8(tmpbuf, (UV)c); |
784 | return is_utf8_print(tmpbuf); |
785 | } |
786 | |
b8c5462f |
787 | bool |
f248d071 |
788 | Perl_is_uni_punct(pTHX_ U32 c) |
b8c5462f |
789 | { |
ad391ad9 |
790 | U8 tmpbuf[UTF8_MAXLEN+1]; |
b8c5462f |
791 | uv_to_utf8(tmpbuf, (UV)c); |
792 | return is_utf8_punct(tmpbuf); |
793 | } |
794 | |
4d61ec05 |
795 | bool |
796 | Perl_is_uni_xdigit(pTHX_ U32 c) |
797 | { |
ad391ad9 |
798 | U8 tmpbuf[UTF8_MAXLEN+1]; |
4d61ec05 |
799 | uv_to_utf8(tmpbuf, (UV)c); |
800 | return is_utf8_xdigit(tmpbuf); |
801 | } |
802 | |
a0ed51b3 |
803 | U32 |
864dbfa3 |
804 | Perl_to_uni_upper(pTHX_ U32 c) |
a0ed51b3 |
805 | { |
ad391ad9 |
806 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
807 | uv_to_utf8(tmpbuf, (UV)c); |
808 | return to_utf8_upper(tmpbuf); |
809 | } |
810 | |
811 | U32 |
864dbfa3 |
812 | Perl_to_uni_title(pTHX_ U32 c) |
a0ed51b3 |
813 | { |
ad391ad9 |
814 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
815 | uv_to_utf8(tmpbuf, (UV)c); |
816 | return to_utf8_title(tmpbuf); |
817 | } |
818 | |
819 | U32 |
864dbfa3 |
820 | Perl_to_uni_lower(pTHX_ U32 c) |
a0ed51b3 |
821 | { |
ad391ad9 |
822 | U8 tmpbuf[UTF8_MAXLEN+1]; |
a0ed51b3 |
823 | uv_to_utf8(tmpbuf, (UV)c); |
824 | return to_utf8_lower(tmpbuf); |
825 | } |
826 | |
827 | /* for now these all assume no locale info available for Unicode > 255 */ |
828 | |
829 | bool |
864dbfa3 |
830 | Perl_is_uni_alnum_lc(pTHX_ U32 c) |
a0ed51b3 |
831 | { |
832 | return is_uni_alnum(c); /* XXX no locale support yet */ |
833 | } |
834 | |
835 | bool |
b8c5462f |
836 | Perl_is_uni_alnumc_lc(pTHX_ U32 c) |
837 | { |
838 | return is_uni_alnumc(c); /* XXX no locale support yet */ |
839 | } |
840 | |
841 | bool |
864dbfa3 |
842 | Perl_is_uni_idfirst_lc(pTHX_ U32 c) |
a0ed51b3 |
843 | { |
844 | return is_uni_idfirst(c); /* XXX no locale support yet */ |
845 | } |
846 | |
847 | bool |
864dbfa3 |
848 | Perl_is_uni_alpha_lc(pTHX_ U32 c) |
a0ed51b3 |
849 | { |
850 | return is_uni_alpha(c); /* XXX no locale support yet */ |
851 | } |
852 | |
853 | bool |
4d61ec05 |
854 | Perl_is_uni_ascii_lc(pTHX_ U32 c) |
855 | { |
856 | return is_uni_ascii(c); /* XXX no locale support yet */ |
857 | } |
858 | |
859 | bool |
864dbfa3 |
860 | Perl_is_uni_space_lc(pTHX_ U32 c) |
a0ed51b3 |
861 | { |
862 | return is_uni_space(c); /* XXX no locale support yet */ |
863 | } |
864 | |
865 | bool |
864dbfa3 |
866 | Perl_is_uni_digit_lc(pTHX_ U32 c) |
a0ed51b3 |
867 | { |
868 | return is_uni_digit(c); /* XXX no locale support yet */ |
869 | } |
870 | |
871 | bool |
864dbfa3 |
872 | Perl_is_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
873 | { |
874 | return is_uni_upper(c); /* XXX no locale support yet */ |
875 | } |
876 | |
877 | bool |
864dbfa3 |
878 | Perl_is_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
879 | { |
880 | return is_uni_lower(c); /* XXX no locale support yet */ |
881 | } |
882 | |
883 | bool |
b8c5462f |
884 | Perl_is_uni_cntrl_lc(pTHX_ U32 c) |
885 | { |
886 | return is_uni_cntrl(c); /* XXX no locale support yet */ |
887 | } |
888 | |
889 | bool |
890 | Perl_is_uni_graph_lc(pTHX_ U32 c) |
891 | { |
892 | return is_uni_graph(c); /* XXX no locale support yet */ |
893 | } |
894 | |
895 | bool |
864dbfa3 |
896 | Perl_is_uni_print_lc(pTHX_ U32 c) |
a0ed51b3 |
897 | { |
898 | return is_uni_print(c); /* XXX no locale support yet */ |
899 | } |
900 | |
b8c5462f |
901 | bool |
902 | Perl_is_uni_punct_lc(pTHX_ U32 c) |
903 | { |
904 | return is_uni_punct(c); /* XXX no locale support yet */ |
905 | } |
906 | |
4d61ec05 |
907 | bool |
908 | Perl_is_uni_xdigit_lc(pTHX_ U32 c) |
909 | { |
910 | return is_uni_xdigit(c); /* XXX no locale support yet */ |
911 | } |
912 | |
a0ed51b3 |
913 | U32 |
864dbfa3 |
914 | Perl_to_uni_upper_lc(pTHX_ U32 c) |
a0ed51b3 |
915 | { |
916 | return to_uni_upper(c); /* XXX no locale support yet */ |
917 | } |
918 | |
919 | U32 |
864dbfa3 |
920 | Perl_to_uni_title_lc(pTHX_ U32 c) |
a0ed51b3 |
921 | { |
922 | return to_uni_title(c); /* XXX no locale support yet */ |
923 | } |
924 | |
925 | U32 |
864dbfa3 |
926 | Perl_to_uni_lower_lc(pTHX_ U32 c) |
a0ed51b3 |
927 | { |
928 | return to_uni_lower(c); /* XXX no locale support yet */ |
929 | } |
930 | |
a0ed51b3 |
931 | bool |
864dbfa3 |
932 | Perl_is_utf8_alnum(pTHX_ U8 *p) |
a0ed51b3 |
933 | { |
386d01d6 |
934 | if (!is_utf8_char(p)) |
935 | return FALSE; |
a0ed51b3 |
936 | if (!PL_utf8_alnum) |
289d4f09 |
937 | /* NOTE: "IsWord", not "IsAlnum", since Alnum is a true |
938 | * descendant of isalnum(3), in other words, it doesn't |
939 | * contain the '_'. --jhi */ |
940 | PL_utf8_alnum = swash_init("utf8", "IsWord", &PL_sv_undef, 0, 0); |
a0ed51b3 |
941 | return swash_fetch(PL_utf8_alnum, p); |
942 | /* return *p == '_' || is_utf8_alpha(p) || is_utf8_digit(p); */ |
943 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
944 | if (!PL_utf8_alnum) |
945 | PL_utf8_alnum = swash_init("utf8", "", |
946 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
947 | return swash_fetch(PL_utf8_alnum, p); |
948 | #endif |
949 | } |
950 | |
951 | bool |
b8c5462f |
952 | Perl_is_utf8_alnumc(pTHX_ U8 *p) |
953 | { |
386d01d6 |
954 | if (!is_utf8_char(p)) |
955 | return FALSE; |
b8c5462f |
956 | if (!PL_utf8_alnum) |
957 | PL_utf8_alnum = swash_init("utf8", "IsAlnumC", &PL_sv_undef, 0, 0); |
958 | return swash_fetch(PL_utf8_alnum, p); |
959 | /* return is_utf8_alpha(p) || is_utf8_digit(p); */ |
960 | #ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */ |
961 | if (!PL_utf8_alnum) |
962 | PL_utf8_alnum = swash_init("utf8", "", |
963 | sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0); |
964 | return swash_fetch(PL_utf8_alnum, p); |
965 | #endif |
966 | } |
967 | |
968 | bool |
864dbfa3 |
969 | Perl_is_utf8_idfirst(pTHX_ U8 *p) |
a0ed51b3 |
970 | { |
971 | return *p == '_' || is_utf8_alpha(p); |
972 | } |
973 | |
974 | bool |
864dbfa3 |
975 | Perl_is_utf8_alpha(pTHX_ U8 *p) |
a0ed51b3 |
976 | { |
386d01d6 |
977 | if (!is_utf8_char(p)) |
978 | return FALSE; |
a0ed51b3 |
979 | if (!PL_utf8_alpha) |
e24b16f9 |
980 | PL_utf8_alpha = swash_init("utf8", "IsAlpha", &PL_sv_undef, 0, 0); |
a0ed51b3 |
981 | return swash_fetch(PL_utf8_alpha, p); |
982 | } |
983 | |
984 | bool |
b8c5462f |
985 | Perl_is_utf8_ascii(pTHX_ U8 *p) |
986 | { |
386d01d6 |
987 | if (!is_utf8_char(p)) |
988 | return FALSE; |
b8c5462f |
989 | if (!PL_utf8_ascii) |
990 | PL_utf8_ascii = swash_init("utf8", "IsAscii", &PL_sv_undef, 0, 0); |
991 | return swash_fetch(PL_utf8_ascii, p); |
992 | } |
993 | |
994 | bool |
864dbfa3 |
995 | Perl_is_utf8_space(pTHX_ U8 *p) |
a0ed51b3 |
996 | { |
386d01d6 |
997 | if (!is_utf8_char(p)) |
998 | return FALSE; |
a0ed51b3 |
999 | if (!PL_utf8_space) |
3bec3564 |
1000 | PL_utf8_space = swash_init("utf8", "IsSpacePerl", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1001 | return swash_fetch(PL_utf8_space, p); |
1002 | } |
1003 | |
1004 | bool |
864dbfa3 |
1005 | Perl_is_utf8_digit(pTHX_ U8 *p) |
a0ed51b3 |
1006 | { |
386d01d6 |
1007 | if (!is_utf8_char(p)) |
1008 | return FALSE; |
a0ed51b3 |
1009 | if (!PL_utf8_digit) |
e24b16f9 |
1010 | PL_utf8_digit = swash_init("utf8", "IsDigit", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1011 | return swash_fetch(PL_utf8_digit, p); |
1012 | } |
1013 | |
1014 | bool |
864dbfa3 |
1015 | Perl_is_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
1016 | { |
386d01d6 |
1017 | if (!is_utf8_char(p)) |
1018 | return FALSE; |
a0ed51b3 |
1019 | if (!PL_utf8_upper) |
e24b16f9 |
1020 | PL_utf8_upper = swash_init("utf8", "IsUpper", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1021 | return swash_fetch(PL_utf8_upper, p); |
1022 | } |
1023 | |
1024 | bool |
864dbfa3 |
1025 | Perl_is_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
1026 | { |
386d01d6 |
1027 | if (!is_utf8_char(p)) |
1028 | return FALSE; |
a0ed51b3 |
1029 | if (!PL_utf8_lower) |
e24b16f9 |
1030 | PL_utf8_lower = swash_init("utf8", "IsLower", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1031 | return swash_fetch(PL_utf8_lower, p); |
1032 | } |
1033 | |
1034 | bool |
b8c5462f |
1035 | Perl_is_utf8_cntrl(pTHX_ U8 *p) |
1036 | { |
386d01d6 |
1037 | if (!is_utf8_char(p)) |
1038 | return FALSE; |
b8c5462f |
1039 | if (!PL_utf8_cntrl) |
1040 | PL_utf8_cntrl = swash_init("utf8", "IsCntrl", &PL_sv_undef, 0, 0); |
1041 | return swash_fetch(PL_utf8_cntrl, p); |
1042 | } |
1043 | |
1044 | bool |
1045 | Perl_is_utf8_graph(pTHX_ U8 *p) |
1046 | { |
386d01d6 |
1047 | if (!is_utf8_char(p)) |
1048 | return FALSE; |
b8c5462f |
1049 | if (!PL_utf8_graph) |
1050 | PL_utf8_graph = swash_init("utf8", "IsGraph", &PL_sv_undef, 0, 0); |
1051 | return swash_fetch(PL_utf8_graph, p); |
1052 | } |
1053 | |
1054 | bool |
864dbfa3 |
1055 | Perl_is_utf8_print(pTHX_ U8 *p) |
a0ed51b3 |
1056 | { |
386d01d6 |
1057 | if (!is_utf8_char(p)) |
1058 | return FALSE; |
a0ed51b3 |
1059 | if (!PL_utf8_print) |
e24b16f9 |
1060 | PL_utf8_print = swash_init("utf8", "IsPrint", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1061 | return swash_fetch(PL_utf8_print, p); |
1062 | } |
1063 | |
1064 | bool |
b8c5462f |
1065 | Perl_is_utf8_punct(pTHX_ U8 *p) |
1066 | { |
386d01d6 |
1067 | if (!is_utf8_char(p)) |
1068 | return FALSE; |
b8c5462f |
1069 | if (!PL_utf8_punct) |
1070 | PL_utf8_punct = swash_init("utf8", "IsPunct", &PL_sv_undef, 0, 0); |
1071 | return swash_fetch(PL_utf8_punct, p); |
1072 | } |
1073 | |
1074 | bool |
1075 | Perl_is_utf8_xdigit(pTHX_ U8 *p) |
1076 | { |
386d01d6 |
1077 | if (!is_utf8_char(p)) |
1078 | return FALSE; |
b8c5462f |
1079 | if (!PL_utf8_xdigit) |
1080 | PL_utf8_xdigit = swash_init("utf8", "IsXDigit", &PL_sv_undef, 0, 0); |
1081 | return swash_fetch(PL_utf8_xdigit, p); |
1082 | } |
1083 | |
1084 | bool |
864dbfa3 |
1085 | Perl_is_utf8_mark(pTHX_ U8 *p) |
a0ed51b3 |
1086 | { |
386d01d6 |
1087 | if (!is_utf8_char(p)) |
1088 | return FALSE; |
a0ed51b3 |
1089 | if (!PL_utf8_mark) |
e24b16f9 |
1090 | PL_utf8_mark = swash_init("utf8", "IsM", &PL_sv_undef, 0, 0); |
a0ed51b3 |
1091 | return swash_fetch(PL_utf8_mark, p); |
1092 | } |
1093 | |
2104c8d9 |
1094 | UV |
864dbfa3 |
1095 | Perl_to_utf8_upper(pTHX_ U8 *p) |
a0ed51b3 |
1096 | { |
1097 | UV uv; |
1098 | |
1099 | if (!PL_utf8_toupper) |
e24b16f9 |
1100 | PL_utf8_toupper = swash_init("utf8", "ToUpper", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1101 | uv = swash_fetch(PL_utf8_toupper, p); |
756820e3 |
1102 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1103 | } |
1104 | |
2104c8d9 |
1105 | UV |
864dbfa3 |
1106 | Perl_to_utf8_title(pTHX_ U8 *p) |
a0ed51b3 |
1107 | { |
1108 | UV uv; |
1109 | |
1110 | if (!PL_utf8_totitle) |
e24b16f9 |
1111 | PL_utf8_totitle = swash_init("utf8", "ToTitle", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1112 | uv = swash_fetch(PL_utf8_totitle, p); |
756820e3 |
1113 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1114 | } |
1115 | |
2104c8d9 |
1116 | UV |
864dbfa3 |
1117 | Perl_to_utf8_lower(pTHX_ U8 *p) |
a0ed51b3 |
1118 | { |
1119 | UV uv; |
1120 | |
1121 | if (!PL_utf8_tolower) |
e24b16f9 |
1122 | PL_utf8_tolower = swash_init("utf8", "ToLower", &PL_sv_undef, 4, 0); |
a0ed51b3 |
1123 | uv = swash_fetch(PL_utf8_tolower, p); |
756820e3 |
1124 | return uv ? uv : utf8_to_uv(p,UTF8_MAXLEN,0,0); |
a0ed51b3 |
1125 | } |
1126 | |
1127 | /* a "swash" is a swatch hash */ |
1128 | |
1129 | SV* |
864dbfa3 |
1130 | Perl_swash_init(pTHX_ char* pkg, char* name, SV *listsv, I32 minbits, I32 none) |
a0ed51b3 |
1131 | { |
1132 | SV* retval; |
bf1fed83 |
1133 | SV* tokenbufsv = sv_2mortal(NEWSV(0,0)); |
8e84507e |
1134 | dSP; |
ce3b816e |
1135 | |
1136 | if (!gv_stashpv(pkg, 0)) { /* demand load utf8 */ |
1137 | ENTER; |
1138 | Perl_load_module(aTHX_ PERL_LOADMOD_NOIMPORT, newSVpv(pkg,0), Nullsv); |
1139 | LEAVE; |
1140 | } |
1141 | SPAGAIN; |
a0ed51b3 |
1142 | PUSHSTACKi(PERLSI_MAGIC); |
1143 | PUSHMARK(SP); |
1144 | EXTEND(SP,5); |
1145 | PUSHs(sv_2mortal(newSVpvn(pkg, strlen(pkg)))); |
1146 | PUSHs(sv_2mortal(newSVpvn(name, strlen(name)))); |
1147 | PUSHs(listsv); |
1148 | PUSHs(sv_2mortal(newSViv(minbits))); |
1149 | PUSHs(sv_2mortal(newSViv(none))); |
1150 | PUTBACK; |
1151 | ENTER; |
1152 | SAVEI32(PL_hints); |
1153 | PL_hints = 0; |
1154 | save_re_context(); |
bf1fed83 |
1155 | if (PL_curcop == &PL_compiling) |
1156 | /* XXX ought to be handled by lex_start */ |
1157 | sv_setpv(tokenbufsv, PL_tokenbuf); |
864dbfa3 |
1158 | if (call_method("SWASHNEW", G_SCALAR)) |
8e84507e |
1159 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
1160 | else |
e24b16f9 |
1161 | retval = &PL_sv_undef; |
a0ed51b3 |
1162 | LEAVE; |
1163 | POPSTACK; |
e24b16f9 |
1164 | if (PL_curcop == &PL_compiling) { |
bf1fed83 |
1165 | STRLEN len; |
1166 | char* pv = SvPV(tokenbufsv, len); |
1167 | |
1168 | Copy(pv, PL_tokenbuf, len+1, char); |
e24b16f9 |
1169 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
1170 | } |
1171 | if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV) |
cea2e8a9 |
1172 | Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref"); |
a0ed51b3 |
1173 | return retval; |
1174 | } |
1175 | |
1176 | UV |
864dbfa3 |
1177 | Perl_swash_fetch(pTHX_ SV *sv, U8 *ptr) |
a0ed51b3 |
1178 | { |
1179 | HV* hv = (HV*)SvRV(sv); |
1180 | U32 klen = UTF8SKIP(ptr) - 1; |
1181 | U32 off = ptr[klen] & 127; /* NB: 64 bit always 0 when len > 1 */ |
1182 | STRLEN slen; |
1183 | STRLEN needents = (klen ? 64 : 128); |
dfe13c55 |
1184 | U8 *tmps; |
a0ed51b3 |
1185 | U32 bit; |
1186 | SV *retval; |
1187 | |
1188 | /* |
1189 | * This single-entry cache saves about 1/3 of the utf8 overhead in test |
1190 | * suite. (That is, only 7-8% overall over just a hash cache. Still, |
1191 | * it's nothing to sniff at.) Pity we usually come through at least |
1192 | * two function calls to get here... |
1193 | * |
1194 | * NB: this code assumes that swatches are never modified, once generated! |
1195 | */ |
1196 | |
1197 | if (hv == PL_last_swash_hv && |
1198 | klen == PL_last_swash_klen && |
12ae5dfc |
1199 | (!klen || memEQ((char *)ptr,(char *)PL_last_swash_key,klen)) ) |
a0ed51b3 |
1200 | { |
1201 | tmps = PL_last_swash_tmps; |
1202 | slen = PL_last_swash_slen; |
1203 | } |
1204 | else { |
1205 | /* Try our second-level swatch cache, kept in a hash. */ |
dfe13c55 |
1206 | SV** svp = hv_fetch(hv, (char*)ptr, klen, FALSE); |
a0ed51b3 |
1207 | |
1208 | /* If not cached, generate it via utf8::SWASHGET */ |
dfe13c55 |
1209 | if (!svp || !SvPOK(*svp) || !(tmps = (U8*)SvPV(*svp, slen))) { |
a0ed51b3 |
1210 | dSP; |
1211 | ENTER; |
1212 | SAVETMPS; |
1213 | save_re_context(); |
1214 | PUSHSTACKi(PERLSI_MAGIC); |
1215 | PUSHMARK(SP); |
1216 | EXTEND(SP,3); |
1217 | PUSHs((SV*)sv); |
756820e3 |
1218 | PUSHs(sv_2mortal(newSViv(utf8_to_uv(ptr, UTF8_MAXLEN, 0, 0) & ~(needents - 1)))); |
a0ed51b3 |
1219 | PUSHs(sv_2mortal(newSViv(needents))); |
1220 | PUTBACK; |
864dbfa3 |
1221 | if (call_method("SWASHGET", G_SCALAR)) |
8e84507e |
1222 | retval = newSVsv(*PL_stack_sp--); |
a0ed51b3 |
1223 | else |
e24b16f9 |
1224 | retval = &PL_sv_undef; |
a0ed51b3 |
1225 | POPSTACK; |
1226 | FREETMPS; |
1227 | LEAVE; |
e24b16f9 |
1228 | if (PL_curcop == &PL_compiling) |
1229 | PL_curcop->op_private = PL_hints; |
a0ed51b3 |
1230 | |
dfe13c55 |
1231 | svp = hv_store(hv, (char*)ptr, klen, retval, 0); |
a0ed51b3 |
1232 | |
dfe13c55 |
1233 | if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) || slen < 8) |
cea2e8a9 |
1234 | Perl_croak(aTHX_ "SWASHGET didn't return result of proper length"); |
a0ed51b3 |
1235 | } |
1236 | |
1237 | PL_last_swash_hv = hv; |
1238 | PL_last_swash_klen = klen; |
1239 | PL_last_swash_tmps = tmps; |
1240 | PL_last_swash_slen = slen; |
1241 | if (klen) |
1242 | Copy(ptr, PL_last_swash_key, klen, U8); |
1243 | } |
1244 | |
9faf8d75 |
1245 | switch ((int)((slen << 3) / needents)) { |
a0ed51b3 |
1246 | case 1: |
1247 | bit = 1 << (off & 7); |
1248 | off >>= 3; |
1249 | return (tmps[off] & bit) != 0; |
1250 | case 8: |
1251 | return tmps[off]; |
1252 | case 16: |
1253 | off <<= 1; |
1254 | return (tmps[off] << 8) + tmps[off + 1] ; |
1255 | case 32: |
1256 | off <<= 2; |
1257 | return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ; |
1258 | } |
cea2e8a9 |
1259 | Perl_croak(aTHX_ "panic: swash_fetch"); |
a0ed51b3 |
1260 | return 0; |
1261 | } |